Part 3 Infinite series, infinite products, and infinite fractions CHAPTER 5 Advanced theory of infinite series Even as the finite encloses an infinite series And in the unlimited limits appear, So the soul of immensity dwells in minutia And in the narrowest limits no limit in here. What joy to discern the minute in infinity! The vast to perceive in the small, what divinity! Jacob Bernoulli (1654-1705) Ars Conjectandi. This chapter is about going in-depth into the theory and application of infinite series. One infinite series that will come up again and again in this chapter and the next chapter as well, is the Riemann zeta function 1 1 ζ(z) = ; nz n=1 X introduced in Section 4.6. Amongst many other things, in this chapter we'll see how to write some well-known constants in terms of the Riemann zeta function; e.g. we'll derive the following neat formula for our friend log 2 ( 5.5): x 1 1 log 2 = ζ(n); 2n n=2 X another formula for our friend the Euler-Mascheroni constant ( 5.9): x 1 ( 1)n γ = − ζ(n); n n=2 X and two more formulas involving our most delicious friend π (see 's 5.10 and 5.11): x 1 3n 1 π2 1 1 1 1 1 π = − ζ(n + 1) ; = ζ(2) = = 1 + + + + : 4n 6 n2 22 32 42 · · · n=2 n=1 X X In this chapter, we'll also derive Gregory-Leibniz-Madhava's formula ( 5.10) x π 1 1 1 1 1 = 1 + + + ; 4 − 3 5 − 7 9 − 11 − · · · and Machin's formula which started the \decimal place race" of computing π ( 5.10): x 1 1 1 ( 1)n 4 1 π = 4 arctan arctan = 4 − : 5 − 239 (2n + 1) 52n+1 − 2392n+1 n=0 X 229 230 5. ADVANCED THEORY OF INFINITE SERIES Leibniz's formula for π=4 is an example of an \alternating series". We study these types of series in Section 5.1. In Section 5.2 and Section 5.3 we look at the ratio and root tests, which you are probably familiar with from elementary calculus. In Section 5.4 we look at power series and prove some pretty powerful properties 3n 1 1 of power series. The formula for log 2, γ, and the formula π = n=2 4−n ζ(n + 1) displayed above are proved using a famous theorem called the Cauchy double series theorem. This theorem, and double sequences and series in general,P are the subject of Section 5.5. In Section 5.6 we investigate rearranging (that is, mixing up the order of the terms in a) series. Here's an interesting question: Does the series 1 1 1 1 1 1 1 1 1 1 1 = + + + + + + + + + + p 2 3 5 7 11 13 17 19 23 29 · · · p isXprime converge or diverge? For the answer, see Section 5.7. In elementary calculus, you probably never seen the power series representations of tangent and secant. This is because these series are somewhat sophisticated mathematically speaking. In Section 5.8 we shall derive the power representations 2n 2n 1 n 1 2 (2 1) B2n 2n 1 tan z = ( 1) − − z − ; − (2n)! n=1 X and 1 E sec z = ( 1)n 2n z2n: − (2n)! n=0 X Here, the B2n's are called \Bernoulli numbers" and the E2n's are called \Euler numbers," which are certain numbers having extraordinary properties. Although you've probably never seen the tangent and secant power series, you might have seen the logarithmic, binomial, and arctangent series: n 1 2n+1 1 ( 1) − 1 α 1 z log(1 + z) = − zn ; (1 + z)α = zn ; arctan z = ( 1)n n n − 2n + 1 n=1 n=0 n=0 X X X where α R. You most likely used calculus (derivatives and integrals) to derive these form2ulæ. In Section 5.9 we shall derive these formulæ without any calculus. Finally, in Sections 5.10 and 5.11 we derive many incredible and awe-inspiring formulæ involving π. In particular, in Section 5.11 we look at the famous Basel problem: Find the sum of the reciprocals of the squares of the natural numbers, 1 2 1 n=1 n2 . The answer, first given by Euler in 1734, is π =6. Chapter 5 objectives: The student will be able to P : : : determine the convergence, and radius and interval of convergence, for an infinite • series and power series, respectively, using various tests, e.g. Dirichlet, Abel, ratio, root, and others. apply Cauchy's double series theorem and know how it relates to rearrangement, • and multiplication and composition of power series. identify series formulæ for the various elementary functions (binomial, arctan- • gent, etc.) and for π. 5.1. SUMMATION BY PARTS, BOUNDED VARIATION, AND ALTERNATING SERIES 231 5.1. Summation by parts, bounded variation, and alternating series In elementary calculus, you studied \integration by parts," a formula I'm sure you used quite often trying to integrate tricky integrals. In this section we study a discrete version of the integration by parts formula called \summation by parts," which is used to sum tricky summations! Summation by parts has broad appli- cations, including finding sums of powers of integers and to derive some famous convergence tests for series, the Dirichlet and Abel tests. 5.1.1. Summation by parts and Abel's lemma. Here is the famous sum- mation by parts formula. The formula is complicated, but the proof is simple. Theorem 5.1 (Summation by parts). For any complex sequences an and b , we have f g f ng n n b (a a ) + a (b b ) = a b a b : k+1 k+1 − k k k+1 − k n+1 n+1 − m m kX=m kX=m Proof. Combining the two terms on the left, we obtain n n b a b a + a b a b = b a a b ; k+1 k+1 − k+1 k k k+1 − k k k+1 k+1 − k k k=m k=m X h i X which is a telescoping sum, leaving us with only a b a b . n+1 n+1 − m m As an easy corollary, we get Abel's lemma named after Niels Abel1 (1802{ 1829). Corollary 5.2 (Abel's lemma). Let a and b be any complex se- f ng f ng quences and let sn denote the n-th partial sum of the series corresponding to the sequence a . Then for any m < n we have f ng n n 1 − a b = s b s b s (b b ): k k n n − m m − k k+1 − k k=Xm+1 kX=m Proof. Applying the summation by parts formula to the sequences sn and b , we obtain f g f ng n 1 n 1 − − b (s s ) + s (b b ) = s b s b : k+1 k+1 − k k k+1 − k n n − m m kX=m kX=m Since a = s s , we conclude that k+1 k+1 − k n 1 n 1 − − b a + s (b b ) = s b s b : k+1 k+1 k k+1 − k n n − m m kX=m kX=m Replacing k with k 1 in the first sum and bringing the second sum to the right, we get our result. − Summation by parts is a very useful tool. We shall apply it find sums of powers of integers (cf. [189], [60]); see the exercises for more applications. 1Abel has left mathematicians enough to keep them busy for 500 years. Charles Hermite (1822{1901), in \Calculus Gems" [161]. 232 5. ADVANCED THEORY OF INFINITE SERIES 5.1.2. Sums of powers of integers. Example 5.1. Let ak = k and bk = k. Then each of the differences ak+1 ak and b b equals 1, so by summation by parts, we have − k+1 − k n n (k + 1) + k = (n + 1)(n + 1) 1 1: − · kX=1 kX=1 This sum reduces to n 2 k = (n + 1)2 n 1 = n(n + 1); − − kX=1 which gives the well-known result: n(n + 1) 1 + 2 + + n = : · · · 2 Example 5.2. Now let a = k2 k = k(k 1) and b = k 1=2. In this case, k − − k − ak+1 ak = (k + 1)k k(k 1) = 2k and bk+1 bk = 1, so by the summation by parts−formula, − − − n 1 n 1 k + (2k) + (k2 k)(1) = (n + 1)n n + : 2 − · 2 k=1 k=1 X X n The first sum on the left contains the sum k=1 k and the second one contains the negative of the same sum. Cancelling, we get P n n(n + 1)(2n + 1) 3 k2 = ; 2 kX=1 which gives the well-known result: n(n + 1)(2n + 1) 12 + 22 + + n2 = : · · · 6 2 2 Example 5.3. For our final result, let ak = k and bk = (k 1) . Then 2 2 − ak+1 ak = (k + 1) k = 2k + 1 and bk+1 bk = 2k 1, so by the summation by parts− formula, − − − n n (k + 1)2(2k + 1) + k2(2k 1) = (n + 1)2 n2: − · kX=1 kX=1 Simplifying the left-hand side, we get n2(n + 1)2 13 + 23 + + n3 = : · · · 4 In Section 12.5, using integration techniques, we'll find a formula for 1p+ +np for any natural number p in terms of the Bernoulli numbers.